pith. sign in

arxiv: 2606.20245 · v1 · pith:YRJFVKMYnew · submitted 2026-06-18 · 💻 cs.AI

Navigating Unreliable Parametric and Contextual Knowledge: Explicit Knowledge Conflict Resolution for LLM Inference

Huang Peng , Jiuyang Tang , Weixin Zeng , Hao Xu , Xiang Zhao This is my paper

Pith reviewed 2026-06-26 17:14 UTC · model grok-4.3

classification 💻 cs.AI
keywords knowledge conflict resolutionlarge language modelsmulti-agent reasoningsemantic entropyparametric knowledgecontextual knowledgeLLM inference
0
0 comments X

The pith

MACR resolves LLM knowledge conflicts by assessing confidence and using multi-agent reasoning to reconcile internal and external sources instead of choosing one.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Large language models draw on both parametric knowledge stored in weights and contextual knowledge supplied in prompts, yet these sources can conflict with each other or internally. Existing methods typically avoid the problem by assuming one source is reliable and discarding the other. The paper introduces MACR, which first uses a modified semantic entropy measure to gauge the model's confidence and decide whether to externalize internal knowledge or retrieve external knowledge. It then deploys three specialized agents to induce explicit rules from the contexts, detect conflicts, and produce a resolved answer. A sympathetic reader would care because reliable handling of mixed, potentially erroneous knowledge sources is required for trustworthy LLM use in real applications.

Core claim

The paper claims that an adaptive knowledge assessment step based on modified semantic entropy, followed by an inductive multi-agent reasoning framework with three agents that respectively induce explicit rules, analyze potential conflicts, and resolve inconsistencies, allows LLMs to actively reconcile unreliable parametric and contextual knowledge rather than privileging one source, yielding higher performance on benchmarks and explicit, interpretable conflict resolutions.

What carries the argument

inductive multi-agent reasoning framework with three specialized agents that induce explicit rules, analyze potential conflicts, and resolve inconsistencies across contexts

If this is right

  • MACR can handle cases where both the model's parametric knowledge and the provided contexts contain errors.
  • The method produces interpretable resolutions of explicit conflicts rather than opaque source selection.
  • Performance exceeds state-of-the-art baselines that rely on binary privileging of one knowledge source.
  • The adaptive assessment step decides when internal knowledge suffices and when external retrieval is needed before multi-agent reasoning begins.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The three-agent structure might be extended to additional specialized agents for conflicts involving temporal or domain-specific inconsistencies.
  • Embedding the conflict-resolution step inside retrieval-augmented generation pipelines could reduce downstream hallucinations when retrieved documents contradict the model.
  • The approach may prove especially useful in high-stakes domains where both model weights and retrieved documents are known to be incomplete.

Load-bearing premise

The modified semantic entropy measure accurately quantifies the LLM's true confidence in its parametric answer for the given query, allowing correct decisions on whether to externalize internal knowledge or retrieve external knowledge.

What would settle it

A test set in which the modified semantic entropy indicates high confidence in an incorrect parametric answer, causing the system to discard correct external knowledge and produce an erroneous resolution.

Figures

Figures reproduced from arXiv: 2606.20245 by Hao Xu, Huang Peng, Jiuyang Tang, Weixin Zeng, Xiang Zhao.

Figure 1
Figure 1. Figure 1: An example of knowledge conflict in retrieval-augmented gener [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The framework of proposed method. knowledge (h, r, t0), and the set of external contexts C as input, and outputs the correct entity t ∗ : t ∗ = f((h, r, ?),(h, r, t0), C). (1) To achieve this, the function f must effectively perform a credibility assessment, discerning the correct knowledge while identifying and mitigating the influence of all other incorrect knowledge triples. The ultimate goal is to gene… view at source ↗
Figure 3
Figure 3. Figure 3: The degradation of ROUGE-L scores for each model as the [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
read the original abstract

Large language models (LLMs) have achieved strong performance across a wide range of language-based tasks by leveraging both extensive parametric knowledge and in-context learning ability, enabling them to incorporate external information provided in the input prompt. However, the integration of external knowledge can introduce conflicts, not only between the model's internal parametric knowledge and the external information, but also among multiple pieces of external contexts. Existing approaches typically assume that either the model or the provided context is reliable, overlooking the possibility that both sources may contain errors, and avoid conflicts by privileging one source over the other, rather than actively resolving inconsistencies. To address these limitations, we propose a novel framework MACR for LLM knowledge conflict resolution that moves beyond the conventional binary choice paradigm and incorporates an explicit conflict-resolution mechanism based on a multi-agent reasoning approach. Specifically, we first propose an adaptive knowledge assessment and retrieval approach that employs a modified semantic entropy measure to quantify an LLM's confidence in its answer to a given query. Based on this confidence estimation, MACR either externalizes the model's internal knowledge as textual representations or retrieves relevant external knowledge when internal knowledge is insufficient, generating basic contexts for subsequent reasoning. Then we introduce an inductive multi-agent reasoning framework with three specialized agents that, respectively, induce explicit rules, analyze potential conflicts, and resolve inconsistencies across all available contexts. Empirical results demonstrate that MACR significantly outperforms state-of-the-art baselines across benchmarks, while also providing interpretable resolutions of explicit conflicts.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 1 minor

Summary. The paper proposes MACR, a framework for explicit knowledge conflict resolution in LLMs. It first uses a modified semantic entropy measure in an adaptive assessment step to decide whether to externalize the model's parametric knowledge as text or retrieve external contexts when internal knowledge is deemed insufficient. This generates base contexts that are then processed by an inductive multi-agent reasoning system consisting of three specialized agents (rule induction, conflict analysis, and inconsistency resolution). The central claim is that MACR significantly outperforms state-of-the-art baselines across benchmarks while yielding interpretable explicit conflict resolutions.

Significance. If the performance gains and the reliability of the modified semantic entropy decision step hold under rigorous validation, the work could meaningfully advance LLM inference by shifting from binary privileging of parametric vs. contextual knowledge to active, multi-agent resolution of inconsistencies. The explicit multi-agent structure offers a structured path to interpretability that is absent in many retrieval-augmented or conflict-avoidance baselines.

major comments (1)
  1. [Adaptive knowledge assessment and retrieval approach] Adaptive knowledge assessment section: the modified semantic entropy is presented as the load-bearing mechanism for deciding whether to externalize internal knowledge or retrieve external contexts, yet the manuscript supplies no calibration of this measure against ground-truth correctness of parametric answers, no head-to-head comparison against unmodified semantic entropy or alternative uncertainty estimators, and no ablation on the decision threshold. Without these, the subsequent multi-agent stage may be operating on misclassified contexts, directly affecting both the performance and interpretability claims.
minor comments (1)
  1. The abstract states that MACR 'significantly outperforms state-of-the-art baselines across benchmarks' but does not name the benchmarks or report effect sizes; adding these details would strengthen the summary without altering the technical contribution.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We appreciate the referee's thorough review and constructive criticism. We respond to the major comment as follows and plan to revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Adaptive knowledge assessment and retrieval approach] Adaptive knowledge assessment section: the modified semantic entropy is presented as the load-bearing mechanism for deciding whether to externalize internal knowledge or retrieve external contexts, yet the manuscript supplies no calibration of this measure against ground-truth correctness of parametric answers, no head-to-head comparison against unmodified semantic entropy or alternative uncertainty estimators, and no ablation on the decision threshold. Without these, the subsequent multi-agent stage may be operating on misclassified contexts, directly affecting both the performance and interpretability claims.

    Authors: We concur that the adaptive knowledge assessment using modified semantic entropy requires more rigorous validation to substantiate its role in the framework. The manuscript as submitted does not include calibration against ground-truth, comparisons to unmodified semantic entropy or other estimators, or ablation on the threshold. To address this, we will include in the revision: (1) calibration experiments correlating the entropy measure with actual correctness on datasets with known answers, (2) head-to-head comparisons with standard semantic entropy and alternatives like token-level entropy, and (3) ablation studies varying the decision threshold to show its impact on performance and context selection. These additions will clarify the reliability of the decision step and support the overall claims. revision: yes

Circularity Check

0 steps flagged

No circularity in MACR framework derivation

full rationale

The paper introduces an empirical framework MACR that uses a modified semantic entropy measure for adaptive knowledge assessment followed by multi-agent conflict resolution. No equations, fitted parameters, or self-referential definitions appear in the abstract or description that would reduce any claimed prediction or result to its own inputs by construction. The approach relies on proposing new components and reporting empirical outperformance rather than a closed derivation chain, self-citation load-bearing premise, or ansatz smuggled through prior work. This is the common case of a self-contained empirical proposal without the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

Abstract-only review yields limited visibility into parameters or assumptions; the framework implicitly relies on the existence of a reliable confidence signal and on the ability of prompted agents to perform rule induction and conflict resolution without introducing new inconsistencies.

axioms (2)
  • domain assumption LLMs possess distinct parametric knowledge that can be externalized as text when confidence is high.
    Stated in the problem setup and used to decide between externalization and retrieval.
  • domain assumption Modified semantic entropy provides a faithful scalar measure of the model's answer confidence.
    Central to the adaptive knowledge assessment step.

pith-pipeline@v0.9.1-grok · 5798 in / 1346 out tokens · 25526 ms · 2026-06-26T17:14:39.454170+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

37 extracted references · 24 canonical work pages · 2 internal anchors

  1. [1]

    Retrieval-augmented generation for knowledge-intensive NLP tasks,

    P . Lewis, E. Perez, A. Piktus, F. Petroni, V . Karpukhin, N. Goyal, H. K ¨uttler, M. Lewis, W. Yih, T. Rockt ¨aschel, S. Riedel, and D. Kiela, “Retrieval-augmented generation for knowledge-intensive NLP tasks,” inAdvances in Neural Information Processing Systems 33: Annual Conference on Neural Information Processing Systems 2020, NeurIPS 2020, December 6...

    2020

  2. [2]

    Available: https://proceedings.neurips.cc/paper/ 2020/hash/6b493230205f780e1bc26945df7481e5-Abstract.html

    [Online]. Available: https://proceedings.neurips.cc/paper/ 2020/hash/6b493230205f780e1bc26945df7481e5-Abstract.html

    2020

  3. [3]

    Resolving knowledge conflicts in large language models,

    Y. Wang, S. Feng, H. Wang, W. Shi, V . Balachandran, T. He, and Y. Tsvetkov, “Resolving knowledge conflicts in large language models,”CoRR, vol. abs/2310.00935, 2023. [Online]. Available: https://doi.org/10.48550/arXiv.2310.00935

    work page doi:10.48550/arxiv.2310.00935 2023

  4. [4]

    Knowledge conflicts for LLMs: A survey

    R. Xu, Z. Qi, Z. Guo, C. Wang, H. Wang, Y. Zhang, and W. Xu, “Knowledge conflicts for llms: A survey,” inProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, EMNLP 2024, Miami, FL, USA, November 12-16, 2024, Y. Al-Onaizan, M. Bansal, and Y. Chen, Eds. Association for Computational Linguistics, 2024, pp. 8541–8565. [Onlin...

    work page doi:10.18653/v1/2024.emnlp-main.486 2024

  5. [5]

    Time waits for no one! analysis and challenges of temporal misalignment,

    K. Luu, D. Khashabi, S. Gururangan, K. Mandyam, and N. A. Smith, “Time waits for no one! analysis and challenges of temporal misalignment,” inProceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, NAACL 2022, Seattle, WA, United States, July 10-15, 2022, M. Carpuat, M...

    work page doi:10.18653/v1/2022.naacl-main.435 2022

  6. [6]

    Lost in the Middle: How Language Models Use Long Contexts

    B. Dhingra, J. R. Cole, J. M. Eisenschlos, D. Gillick, J. Eisenstein, and W. W. Cohen, “Time-aware language models as temporal knowledge bases,”Trans. Assoc. Comput. Linguistics, vol. 10, pp. 257–273, 2022. [Online]. Available: https://doi.org/10.1162/tacl\ a\ 00459

    work page internal anchor Pith review doi:10.1162/tacl 2022

  7. [7]

    Attacking open-domain question answering by injecting misinformation,

    L. Pan, W. Chen, M. Kan, and W. Y. Wang, “Attacking open-domain question answering by injecting misinformation,” inProceedings of the 13th International Joint Conference on Natural Language Processing and the 3rd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics, IJCNLP 2023 -Volume 1: Long Papers, Nusa Dua, Bali, Nov...

    work page doi:10.18653/v1/2023.ijcnlp-main.35 2023

  8. [8]

    Who’s who: Large language models meet knowledge conflicts in practice,

    Q. Pham, H. Ngo, A. T. Luu, and D. Q. Nguyen, “Who’s who: Large language models meet knowledge conflicts in practice,” inFindings of the Association for Computational Linguistics: EMNLP 2024, Miami, Florida, USA, November 12-16, 2024, Y. Al-Onaizan, M. Bansal, and Y. Chen, Eds. Association for Computational Linguistics, 2024, pp. 10 142–10 151. [Online]. ...

    work page doi:10.18653/v1/2024.findings-emnlp.593 2024

  9. [9]

    Ai as agency without intelligence: On chatgpt, large language models, and other generative models,

    L. Floridi, “Ai as agency without intelligence: On chatgpt, large language models, and other generative models,”Philosophy & technology, vol. 36, no. 1, p. 15, 2023

    2023

  10. [10]

    Adaptive contrastive decoding in retrieval-augmented generation for handling noisy contexts,

    Y. Kim, H. J. Kim, C. Park, C. Park, H. Cho, J. Kim, K. M. Yoo, S. Lee, and T. Kim, “Adaptive contrastive decoding in retrieval-augmented generation for handling noisy contexts,” inFindings of the Association for Computational Linguistics: EMNLP 2024, Miami, Florida, USA, November 12-16, 2024, Y. Al-Onaizan, M. Bansal, and Y. Chen, Eds. Association for Co...

    work page doi:10.18653/v1/2024.findings-emnlp.136 2024

  11. [11]

    Parameters vs. context: Fine-grained control of knowledge reliance in language models,

    B. Bi, S. Liu, Y. Wang, Y. Xu, J. Fang, L. Mei, and X. Cheng, “Parameters vs. context: Fine-grained control of knowledge reliance in language models,”CoRR, vol. abs/2503.15888, 2025. [Online]. Available: https://doi.org/10.48550/arXiv.2503.15888

    work page doi:10.48550/arxiv.2503.15888 2025

  12. [12]

    Plug-and-play adaptation for continuously-updated QA,

    K. Lee, W. Han, S. Hwang, H. Lee, J. Park, and S. Lee, “Plug-and-play adaptation for continuously-updated QA,” in Findings of the Association for Computational Linguistics: ACL 2022, Dublin, Ireland, May 22-27, 2022, S. Muresan, P . Nakov, and A. Villavicencio, Eds. Association for Computational Linguistics, 2022, pp. 438–447. [Online]. Available: https: ...

    work page doi:10.18653/v1/2022.findings-acl.37 2022

  13. [13]

    V-DPO: Mitigating hallucination in large vision language models via vision-guided direct preference optimization

    Z. Gekhman, J. Herzig, R. Aharoni, C. Elkind, and I. Szpektor, “Trueteacher: Learning factual consistency evaluation with large language models,” inProceedings of the 2023 Conference on JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2021 11 Empirical Methods in Natural Language Processing, EMNLP 2023, Singapore, December 6-10, 2023, H. Bouamor, J. P...

    work page doi:10.18653/v1/ 2023

  14. [14]

    LLM-blender: Ensembling large language models with pairwise ranking and generative fusion

    W. Zhou, S. Zhang, H. Poon, and M. Chen, “Context-faithful prompting for large language models,” inFindings of the Association for Computational Linguistics: EMNLP 2023, Singapore, December 6-10, 2023, H. Bouamor, J. Pino, and K. Bali, Eds. Association for Computational Linguistics, 2023, pp. 14 544– 14 556. [Online]. Available: https://doi.org/10.18653/v...

    work page doi:10.18653/v1/2023 2023

  15. [15]

    Improving factual consistency for knowledge-grounded dialogue systems via knowledge enhancement and alignment,

    B. Xue, W. Wang, H. Wang, F. Mi, R. Wang, Y. Wang, L. Shang, X. Jiang, Q. Liu, and K. Wong, “Improving factual consistency for knowledge-grounded dialogue systems via knowledge enhancement and alignment,” inFindings of the Association for Computational Linguistics: EMNLP 2023, Singapore, December 6-10, 2023, H. Bouamor, J. Pino, and K. Bali, Eds. Associat...

    work page doi:10.18653/v1/2023.findings-emnlp.525 2023

  16. [16]

    Yu and Sanjiv Kumar , editor =

    D. Li, A. S. Rawat, M. Zaheer, X. Wang, M. Lukasik, A. Veit, F. X. Yu, and S. Kumar, “Large language models with controllable working memory,” inFindings of the Association for Computational Linguistics: ACL 2023, Toronto, Canada, July 9-14, 2023, A. Rogers, J. L. Boyd-Graber, and N. Okazaki, Eds. Association for Computational Linguistics, 2023, pp. 1774–...

    work page doi:10.18653/v1/2023.findings-acl.112 2023

  17. [17]

    Trusting Your Evidence: Hallucinate Less with Context-aware Decoding

    W. Shi, X. Han, M. Lewis, Y. Tsvetkov, L. Zettlemoyer, and W. Yih, “Trusting your evidence: Hallucinate less with context-aware decoding,” inProceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies: Short Papers, NAACL 2024, Mexico City, Mexico, June 16-21, 2024, K. Duh,...

    work page doi:10.18653/v1/2024.naacl-short.69 2024

  18. [18]

    IRCAN: mitigating knowledge conflicts in LLM generation via identifying and reweighting context-aware neurons,

    D. Shi, R. Jin, T. Shen, W. Dong, X. Wu, and D. Xiong, “IRCAN: mitigating knowledge conflicts in LLM generation via identifying and reweighting context-aware neurons,” in Advances in Neural Information Processing Systems 38: Annual Conference on Neural Information Processing Systems 2024, NeurIPS 2024, Vancouver, BC, Canada, December 10 - 15, 2024, A. Glo...

    2024

  19. [19]

    Defending against disinformation attacks in open-domain question answering,

    O. Weller, A. Khan, N. Weir, D. J. Lawrie, and B. V . Durme, “Defending against disinformation attacks in open-domain question answering,” inProceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics, EACL 2024 - Volume 2: Short Papers, St. Julian’s, Malta, March 17-22, 2024, Y. Graham and M. Purver, Eds. A...

    2024

  20. [20]

    The earth is flat because...: Investigating llms’ belief towards misinformation via persuasive conversation,

    R. Xu, B. S. Lin, S. Yang, T. Zhang, W. Shi, T. Zhang, Z. Fang, W. Xu, and H. Qiu, “The earth is flat because...: Investigating llms’ belief towards misinformation via persuasive conversation,” inProceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), ACL 2024, Bangkok, Thailand, August 11-16, 2024,...

    work page doi:10.18653/v1/2024.acl-long.858 2024

  21. [21]

    On the Risk of Misinformation Pollution with Large Language Models

    Y. Pan, L. Pan, W. Chen, P . Nakov, M. Kan, and W. Y. Wang, “On the risk of misinformation pollution with large language models,” inFindings of the Association for Computational Linguistics: EMNLP 2023, Singapore, December 6-10, 2023, H. Bouamor, J. Pino, and K. Bali, Eds. Association for Computational Linguistics, 2023, pp. 1389–1403. [Online]. Available...

    work page doi:10.18653/v1/2023.findings-emnlp.97 2023

  22. [22]

    Why so gullible? enhancing the robustness of retrieval- augmented models against counterfactual noise,

    G. Hong, J. Kim, J. Kang, S. Myaeng, and J. J. Whang, “Why so gullible? enhancing the robustness of retrieval- augmented models against counterfactual noise,” inFindings of the Association for Computational Linguistics: NAACL 2024, Mexico City, Mexico, June 16-21, 2024, K. Duh, H. G ´omez- Adorno, and S. Bethard, Eds. Association for Computational Linguis...

    work page doi:10.18653/v1/2024.findings-naacl.159 2024

  23. [23]

    Taming knowledge conflicts in language models,

    G. Li, Y. Chen, and H. Tong, “Taming knowledge conflicts in language models,”CoRR, vol. abs/2503.10996, 2025. [Online]. Available: https://doi.org/10.48550/arXiv.2503.10996

    work page doi:10.48550/arxiv.2503.10996 2025

  24. [24]

    Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval , pages =

    F. Cuconasu, G. Trappolini, F. Siciliano, S. Filice, C. Campagnano, Y. Maarek, N. Tonellotto, and F. Silvestri, “The power of noise: Redefining retrieval for RAG systems,” inProceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, SIGIR 2024, Washington DC, USA, July 14-18, 2024, G. H. Yang, H. Wang,...

    work page doi:10.1145/3626772.3657834 2024

  25. [25]

    Rethinking relevance: How noise and distractors impact retrieval-augmented generation,

    ——, “Rethinking relevance: How noise and distractors impact retrieval-augmented generation,” inProceedings of the 14th Italian Information Retrieval Workshop, Udine, Italy, September 5-6, 2024, ser. CEUR Workshop Proceedings, K. Roitero, M. Viviani, E. Maddalena, and S. Mizzaro, Eds., vol

    2024

  26. [26]

    CEUR-WS.org, 2024, pp. 95–98. [Online]. Available: https://ceur-ws.org/Vol-3802/paper23.pdf

    2024

  27. [27]

    Faithfulrag: Fact-level conflict modeling for context-faithful retrieval-augmented generation,

    Q. Zhang, Z. Xiang, Y. Xiao, L. Wang, J. Li, X. Wang, and J. Su, “Faithfulrag: Fact-level conflict modeling for context-faithful retrieval-augmented generation,” inProceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), ACL 2025, Vienna, Austria, July 27 - August 1, 2025, W. Che, J. Nabende, E. Shut...

    2025

  28. [28]

    Truthfulrag: Resolving factual- level conflicts in retrieval-augmented generation with knowledge graphs,

    S. Liu, Y. Shang, and X. Zhang, “Truthfulrag: Resolving factual- level conflicts in retrieval-augmented generation with knowledge graphs,”CoRR, vol. abs/2511.10375, 2025. [Online]. Available: https://doi.org/10.48550/arXiv.2511.10375

    work page doi:10.48550/arxiv.2511.10375 2025

  29. [29]

    Corrective Retrieval Augmented Generation

    S. Yan, J. Gu, Y. Zhu, and Z. Ling, “Corrective retrieval augmented generation,”CoRR, vol. abs/2401.15884, 2024. [Online]. Available: https://doi.org/10.48550/arXiv.2401.15884

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2401.15884 2024

  30. [30]

    Instructrag: Instructing retrieval- augmented generation via self-synthesized rationales,

    Z. Wei, W. Chen, and Y. Meng, “Instructrag: Instructing retrieval- augmented generation via self-synthesized rationales,” inThe Thirteenth International Conference on Learning Representations, ICLR 2025, Singapore, April 24-28, 2025. OpenReview.net, 2025. [Online]. Available: https://openreview.net/forum?id=P1qhkp8gQT

    2025

  31. [31]

    Truthfulrag: Resolving factual-level conflicts in retrieval-augmented generation with knowledge graphs,

    S. Liu, Y. Shang, and X. Zhang, “Truthfulrag: Resolving factual-level conflicts in retrieval-augmented generation with knowledge graphs,” inFortieth AAAI Conference on Artificial Intelligence, Thirty-Eighth Conference on Innovative Applications of Artificial Intelligence, Sixteenth Symposium on Educational Advances in Artificial Intelligence, AAAI 2026, S...

    work page doi:10.1609/aaai.v40i38.40489 2026

  32. [32]

    Rethinking all evidence: Enhancing trustworthy retrieval-augmented generation via conflict-driven summarization,

    J. Chen, B. Bi, W. Zhang, J. Sui, X. Zhu, Y. Wang, L. Mei, and S. Liu, “Rethinking all evidence: Enhancing trustworthy retrieval-augmented generation via conflict-driven summarization,”CoRR, vol. abs/2507.01281, 2025. [Online]. Available: https://doi.org/10.48550/arXiv.2507.01281

    work page doi:10.48550/arxiv.2507.01281 2025

  33. [33]

    Detecting hallucinations in large language models using semantic entropy , volume =

    S. Farquhar, J. Kossen, L. Kuhn, and Y. Gal, “Detecting hallucinations in large language models using semantic entropy,” Nat., vol. 630, no. 8017, pp. 625–630, 2024. [Online]. Available: https://doi.org/10.1038/s41586-024-07421-0

    work page doi:10.1038/s41586-024-07421-0 2024

  34. [34]

    Beyond semantic entropy: Boosting LLM uncertainty quantification with pairwise semantic similarity,

    D. Nguyen, A. Payani, and B. Mirzasoleiman, “Beyond semantic entropy: Boosting LLM uncertainty quantification with pairwise semantic similarity,” inFindings of the Association for Computational Linguistics, ACL 2025, Vienna, Austria, July 27 - August 1, 2025, W. Che, J. Nabende, E. Shutova, and M. T. Pilehvar, Eds. Association for Computational Linguistic...

    2025

  35. [35]

    Conflictbank: A benchmark for evaluating the influence of knowledge conflicts in llms,

    Z. Su, J. Zhang, X. Qu, T. Zhu, Y. Li, J. Sun, J. Li, M. Zhang, and Y. Cheng, “Conflictbank: A benchmark for evaluating the influence of knowledge conflicts in llms,” inAdvances in Neural Information Processing Systems 38: Annual Conference on Neural Information Processing Systems 2024, NeurIPS 2024, Vancouver, BC, Canada, December 10 - 15, 2024, A. Globe...

    2024

  36. [36]

    Context-dpo: Aligning language models for context- faithfulness,

    B. Bi, S. Huang, Y. Wang, T. Yang, Z. Zhang, H. Huang, L. Mei, J. Fang, Z. Li, F. Wei, W. Deng, F. Sun, Q. Zhang, and S. Liu, “Context-dpo: Aligning language models for context- faithfulness,” inFindings of the Association for Computational Linguistics, ACL 2025, Vienna, Austria, July 27 - August 1, 2025, W. Che, J. Nabende, E. Shutova, and M. T. Pilehvar...

    2025

  37. [37]

    Mquake: Assessing knowledge editing in language models via multi-hop questions,

    Z. Zhong, Z. Wu, C. D. Manning, C. Potts, and D. Chen, “Mquake: Assessing knowledge editing in language models via multi-hop questions,” inProceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, EMNLP 2023, Singapore, December 6-10, 2023, H. Bouamor, J. Pino, and K. Bali, Eds. Association for Computational Linguistics, 2023...

    2023